Sound-Lab A-1 electrostatic loudspeaker Measurements

Sidebar 6: Measurements

To investigate the answer to Dick's question, I measured the A-1's impedance magnitude at three settings of the Brilliance Control (BC) with DRA Labs' MLSSA system. With the BC set at Maximum (pot wide open), I obtained the curves in fig.1. Although the magnitude of the impedance is above 10 ohms from the upper bass to 2kHz, and well above 30 ohms in the bass, note the drop in the mid-treble and above. The cursor indicates a punishingly low 1.3 ohms at 20kHz with an equally punishing phase angle of –75.5°, dropping even further to just a small fraction of an ohm at ultrasonic frequencies. This measurement necessarily includes the resistance of the speaker cables DO was using (TARA Labs Rectangular Solid Core); as the margin of error in my measurement is probably around a fraction of ohm, it's quite possible that the A-1 presents amplifiers with a complete short circuit above the audio range. Fundamentally, therefore, in electrical terms, the A-1 with its BC set to maximum is nothing more than a large capacitor. If this isn't hell for a power amplifier, I don't know what is (footnote 1). No wonder DO's amps were choking.


Fig.1 Sound-Lab A-1, electrical impedance (solid) and phase (dashed) (2 ohms/vertical div.).

Backing off the BC to the Minimum position raised the impedance at 20kHz to a more reasonable 4 ohms. However, this drastically reduced the sonic level of the treble, as can be inferred from fig.2, which shows the effect of the BC referenced to the response with the control set at 12 o'clock (middle curve, which therefore appears as a straight line), which seemed to give the flattest measured response. The control gives a massive maximum response change of 32dB at 23kHz, which is quite excessive, in my opinion.


Fig.2 Sound-Lab A-1, effect of Brilliance Control, normalized to resposne with control set to 12:00 (straight line).

Turning to the time domain, the A-1's impulse response (with the microphone 36" from the ground, about 45" away, and the speaker driven by the Crown Macro Reference amplifier) is shown in fig.3, calculated by MLSSA. (The step response is shown in fig.4.) The impulse response of a perfect single-diaphragm speaker should approximate a band-pass–filtered unidirectional, rectangular pulse (fig.5); the Sound-Lab's impulse response, however, is overlaid with high and complex patterns of high-frequency ringing. This is explored further in the speaker's cumulative spectral decay, or "waterfall," plot (fig.6). For comparison, fig.7 shows how a perfect loudspeaker should look on this kind of plot: a flat initial response decays evenly and cleanly at all frequencies, with no ridges parallel with the time axis indicating the presence of resonances. By contrast, the A-1's diaphragm can be seen to be in a state of severe breakup in the high treble, as well as there being a couple of resonant modes in the low treble.


Fig.3 Sound-Lab A-1, impulse response at 45" with mike 36" from floor (5ms time window, 30kHz bandwidth).


Fig.4 Sound-Lab A-1, step response at 45" with mike 36" from floor (5ms time window, 30kHz bandwidth).


Fig.5 "Perfect loudspeaker," impulse response (5ms time window, 30kHz bandwidth).


Fig.6 Sound-Lab A-1, cumulative spectral decay plot at 45" (0.15ms risetime).


Fig.7 "Perfect loudspeaker," cumulative spectral decay plot at 45" (0.15ms risetime).

There is, I believe, a big misunderstanding over the way in which the diaphragms of planar loudspeakers behave. DO promulgates this misunderstanding when he states above that the A-1 consists of "a single driver driven uniformly over its entire area." That is, of course, true, but it doesn't mean that the diaphragm moves uniformly at all frequencies as a result of that uniform drive. At high frequencies, as can be seen from fig.6, the diaphragm motion is actually chaotic (footnote 2). The average position of the diaphragm may move uniformly in response to the drive signal, but each little element of the diaphragm "shivers" about that average position, with perhaps one element in front of where it should be but its immediate neighbor behind. The trick for the speaker designer is to push this chaotic behavior high enough in frequency that it doesn't intrude on the music. From my own auditioning of the A-1s, I believe Dr. West has managed to do this. The worst behavior in fig.6, for example, is above 15kHz, where the ear's sensitivity falls off rapidly.

Before I look at the A-1's actual frequency response, figs.8 and 9 show the manner in which that response changes for off-axis listeners in both vertical and horizontal planes. As Dick stated, the A-1's design should result in a speaker whose tonal balance doesn't change significantly off-axis. Measuring such large panel speakers relatively close to the diaphragm is an endeavor fraught with difficulty. Nevertheless, the A-1's sound changes quite evenly as the listener moves to the speaker's side (fig.8). (The large peaks and dips are probably interference effects at the 45" microphone position, unnoticeable at a more typical listening position.)


Fig.8 Sound-Lab A-1, horizontal response family at 45", normalized to on-axis response, from back to front: differences in response 90°–5° off-axis; reference response; differences in response 5°–90° off-axis.


Fig.9 Sound-Lab A-1, vertical response family at 45", normalized to on-axis response, from back to front: differences in response 68"–45" from ground; reference response; differences in response 28"–18" from ground.

Note that the frontmost trace in fig.8 is the response 90° off-axis; ie, to the side of the panel. Because I only calculate the changes in response to generate this graph, not the absolute amplitude difference, it can be misleading in that the 90° response is also actually much lower in level than the on-axis response. The dipole A-1 produces a good null to its side, even though the response of the residual signal is not too dissimilar to the main signal.

The A-1's balance doesn't change significantly in the vertical plane (fig.9). (Again, I assume that the peaks and dips are measurement-related interference effects.) This reinforces the idea that such a tall panel in a typical room tends to act as a line source.

The Sound-Lab's frequency response at the 45" measuring distance, averaged across a 30° window, is shown to the right of fig.10. The sloping-down from the lower midrange to the treble is, I believe, due to the proximity effect featured by a large panel—ie, the diaphragm size is significant compared with the microphone distance—and should be ignored. The BC control was set to its maximum; note both the high treble peak and the raggedness of the response in this region. Lower down in frequency, however, the midrange/treble transition region is very smooth.


Fig.10 Sound-Lab A-1, anechoic response at 45" with microphone 36" from floor, averaged across 30° horizontal window and corrected for microphone response, with nearfield response (average of 6 spaced measurements) plotted below 200Hz (Brilliance Control set to maximum).

To the left of fig.10 is the A-1's nearfield response, taken with the microphone almost touching the panel. (Actually, because of Sound-Lab's patented "distributed bass resonance" principle, I took this measurement at six places and averaged the result. The variation was not large, however.) This curve shows the A-1 to roll out below 40Hz; as DO found, without the wings it is not a full-range speaker. What puzzled me, however, was that I was expecting to find a significant peak in the midbass due to the diaphragm's fundamental resonance. For example, to the left of fig.11 can be seen the nearfield responses I took at two different places on the diaphragm of another, older sample of the A-1 owned by a friend of J. Gordon Holt's in Boulder, CO. (To the right is this A-1's response at a 45" distance. It looks smooth because I applied 1/5-octave averaging, but the general shape in the treble is the same as in fig.10.) Dick's nearfield measurements on the first A-1 sample he had auditioned also showed a large peak in the midbass, so something must have changed either in the speaker or in my measurement setup.


Fig.11 Sound-Lab A-1, Colorado sample, 1/5 octave-smoothed anechoic response averaged across 30° horizontal window and corrected for microphone response, with 2 separate nearfield responses plotted below 200Hz.

I was puzzled enough by this result that I went back and remeasured the nearfield low-frequency response of the Wilson Puppy I'd last measured in April 1991. Within a dB here and there, the old and new curves matched, indicating that the B&K 4006 microphone/EAR mike preamplifier I use had not changed in the meantime. Presumably, therefore, this latest sample of the A-1 is behaving in a manner different from that of older versions. (The SALLIE—see later—was in place for these measurements.) Certainly, from my own auditioning, the speaker didn't sound as if it had a peak in the bass, the sound being tight and clean, whereas the Colorado sample did.

I mentioned earlier that I was unable to measure the A-1 when driven by DO's preferred OTL tube amplifier, the Fourier Sans Pareil. As Tom Norton had measured the Fourier's output impedance to accompany the review last June, it was a simple matter to calculate the effect on the A-1's frequency response that would result from the interaction between the speaker's modulus of impedance and the amplifier's relatively high source impedance, which ranges from 0.37 ohms at 1kHz to 0.55 ohms at 20kHz. Fig.12 shows the response between 200Hz and 20kHz from fig.10, but modified with this interaction. Because the speaker's impedance rises to tens of ohms below 200Hz, using the Fourier will boost the bass. In the treble, however, where the speaker's impedance is falling, the response will be increasingly tilted down with the tube amp, with at least a 3dB attenuation at 20kHz, as shown in fig.12.


Fig.12 Sound-Lab A-1, anechoic response above 200Hz from fig. 10, corrected for microphone response and adjusted for interaction between speaker impedance in fig.1 and measured output impedance of Fourier OTL amplifier.

Certainly I found the A-1 driven by the Fourier to offer an extremely natural midrange and treble tonality, with only occasionally a hint of very-high-frequency tizz. But I have to reinforce DO's comments on the A-1's soundstaging, something that will not be revealed by these measurements. The impression of image size was extremely lifelike, much more so than I have experienced from conventional dynamic speakers or even from smaller panel speakers like the Quads.

Finally, figs.13 and 14 show the effect of the wings on the speaker's response, taken by DO at the listening seat. Dick notes that "there is a common conception that an omni mike positioned at the location of the listener's head, together with a swept sinewave response, can be used to accurately predict the perceived sound field. That's just not so.


Fig.13 Sound-Lab A-1, in-room response at DO's listening seat without wings (midrange/bass balance set to –6/+3dB, Brilliance Control set to maximum).


Fig.14 Sound-Lab A-1, in-room response at DO's listening seat with wings (midrange/bass balance set to –6/+3dB, Brilliance Control set to maximum).

"There are at least two major problems with such an approach. While an omni's acceptance pattern does not discriminate against any particular angle of incidence, the pinna and ear canal most definitely do. To quote Jens Blauert, the noted German psychoacoustician: 'The pinna, along with the ear canal, forms a system of acoustical resonators. The degree to which individual resonances of this system are excited depends on the direction and distance of the sound source.' In my experience, the response of a mike with a cardioid pickup pattern correlates more closely with listening impressions.

"Second, there's a problem in using steady-state signals (ie, sinewaves) to measure bass response in a room. Unlike bass transients—which do not fully energize room standing waves—sinewaves build up room modes to their fullest. Thus, a room's resonant signature is more apparent with sinewaves than it is with music signal. Using a warble tone, as I do, to minimize the buildup of room modes, helps in this regard."

With these caveats in mind, fig.13 shows the listening position response with the A-1s' EQ set to –6/+3dB, which DO found to yield the most convincing tonal balance in his room. Ignore the deep dip at 38Hz, which will be due to a specific node at the mike position chosen by DO. With the reinforcement offered by the room, the bass is already quite extended; adding the wings pushes the in-room extension to 20Hz (fig.14), which is why DO decided to try again with them.—John Atkinson

Footnote 1: I had wanted to drive the speaker with Dick's preferred Fourier OTL for these measurements. However, even at RMS levels of 1–4W, the output tube plates glowed cherry red with the low peak:mean ratio MLS signal, and the amp shut down.

Footnote 2: That this motion is mathematically chaotic is revealed by the fact that panel speakers tend to produce subharmonics when driven hard. (The mathematician Manfred Schroeder has said that the production of subharmonics is always an indication of chaotic behavior.) See Stereophile, May 1992, p.109, for example, where I showed the spectrum of a panel speaker reproducing a 1kHz tone at a high level. As well as conventional distortion harmonics that are higher in frequency than the fundamental, the spectrum shows a significant amount of 500Hz being produced by the diaphragm.

Sound-Lab Corporation
153 N 400 W Street
Gunnison, UT 84634
(453) 528-7218

remlab's picture

...issue I ever bought. I never forgot about the Mylar thickness issue. I remember saying to myself, "what if John hadn't tested the speaker?" There would have been a whole bunch of screwed up A-1's out there, probably to this day. That's what made me realize how important the testing of all equipment is. How often has equipment malfunctioned or been out of spec  during Absolute Sound reviews without them even knowing. Obviously, having measured A-1's previously did help in this case.